|Home | About | Journals | Submit | Contact Us | Français|
The functional switch of glutamine/asparagine (Q/N)-rich prions and the neurotoxicity of polyQ-expanded proteins involve complex aggregation-prone structural transitions, commonly presumed to be forming β-sheets. By analyzing sequences of interaction partners of these proteins, we discovered a recurrent presence of coiled-coil domains both in the partners and in segments that flank or overlap Q/N-rich and polyQ domains. Since coiled-coils can mediate protein interactions and multimerization, we studied their possible involvement in Q/N-rich and polyQ aggregations. Using circular dichroism and chemical cross-linking, we found that Q/N-rich and polyQ peptides form α-helical coiled-coils in vitro and assemble into multimers. Using structure-guided mutagenesis, we found that coiled-coil domains modulate in vivo properties of two Q/N-rich prions and polyQ-expanded huntingtin. Mutations that disrupt coiled-coils impair aggregation and activity, whereas mutations that enhance coiled-coil propensity promote aggregation. These findings support a coiled-coil model for the functional switch of Q/N-rich prions and for the pathogenesis of polyQ-expansion diseases.
Prions were initially identified as protein-only infectious agents causing Creutzfeldt-Jakob and related neurological diseases by misfolding from their native conformation to a self-propagating, aggregation-prone, and ß-sheet-rich secondary structure (Prusiner, 1987). Subsequent studies delineated a group of prion proteins in yeast, characterized by glutamine/asparagine-rich (Q/N-rich) domains (Wickner et al., 2007). Some of these proteins are thought to be non-pathogenic and to propagate epigenetic information through a loss of function upon aggregation (Alberti et al., 2009). More recently, a Q/N-rich prion protein has been found that serves a physiological function in the nervous system of Aplysia. Here, the neurotransmitter serotonin controls a prion-like switch that results in a self-perpetuating form that mediates the long-term maintenance of synaptic plasticity and memory storage (Si et al., 2003, 2010). The existence of non-pathogenic and functionally regulated Q/N-rich prions raises questions about the mechanisms by which low-complexity Q/N-rich domains can regulate the physiological function of these proteins in a tightly controlled manner and in response to specific cellular signaling events (Si et al., 2010).
The biochemical properties of Q/N-rich proteins make them rather insoluble and difficult to study by conventional structural techniques. As a result, full atomic-level structures have not yet been determined, and many studies focused instead on peptide fragments (e.g. Perutz et al., 2002; Nelson et al., 2005). Based on such studies, Q/N-rich prions are thought to undergo a conformational change analogous to that of pathogenic prions and amyloids, by misfolding into ß-sheet-rich, aggregating structures (Wickner et al., 2007; Alberti et al., 2009). However, the structural transitions of functional prions, which must be regulated, are not easily conceivable as an uncontrolled misfolding process, suggesting that structural mechanisms besides ß-sheet misfolding might underlie the prion-like behavior of these proteins. Indeed, experimental observations do indicate involvement of other structures (e.g. Bousset et al., 2002; Narayanan et al., 2006).
Since studies of Q/N-rich prions revealed that their structural changes involve chaperones and other interacting proteins (Wickner et al., 2007), we wondered whether identifying common structural features among these interactors might provide clues about the structure and regulation of the Q/N-rich proteins themselves. We therefore undertook an analysis of their interactomes, which revealed the overrepresentation of coiled-coil domains. A similar overrepresentation was found in the interactomes of non-prionic proteins containing polyQ stretches that have a physiological function as the wild-type form (e.g. Harada et al, 2010), but acquire dysfunctional properties when genetic mutations increase their length above a critical threshold where they assume a pathogenic form that causes inherited disorders such as Huntington’s disease (Orr and Zoghbi, 2007).
Coiled-coils (CCs) are α-helical supersecondary structures mediating protein-protein interactions, oligomerization, and other functions through the coiling of helices belonging to the same or different polypeptide chains (Parry et al., 2008). The finding of CC proteins in the interactomes prompted us to search for CC regions in the Q/N-rich and the polyQ proteins themselves, which could serve both as substrates of interaction with other CCs and mediators of aggregation. This analysis revealed heptad repeats typical of CCs in regions flanking or overlapping Q/N-rich domains and polyQ stretches. We used circular dichroism and chemical cross-linking to study properties of Q/N-rich and polyQ-containing peptides in vitro, and found signatures of α-helical CC multimers. To investigate the role of CCs in the aggregation and activity of Q/N-rich and polyQ proteins in vivo, we performed structure-guided mutagenesis of the Q/N-rich yeast prion Ure2, the Aplysia prion CPEB, and the human polyQ protein huntingtin. We found that CC domains regulate aggregation, insolubility and activity of these proteins. Based on these findings, we propose a novel model in which CCs have a critical role in the structural dynamics of Q/N-rich prions and polyQ proteins.
The ubiquitin ligase CHIP, the chaperone Hsp104, the polyQ tract-binding protein-1 (PQBP-1), and the Htt-interacting protein (HIP-1) interact with Q/N-rich or polyQ proteins modulating their aggregation (Orr and Zoghbi, 2007; Wickner et al., 2007). In search of shared features among them, we discovered the common occurrence of CC domains, as previously found in three of them (e.g. Niu and Ybe, 2008), and as predicted with high probability for PQBP-1 (Fig. 1A). We extended our analysis to other interactors of the Q/N-rich prion Ure2 and the polyQ protein Htt listed in the BioGrid database, and found that 54% and 63% of Ure2 and Htt interactors, respectively, have or are predicted to have CC domains (Fig. 1B). The apCPEB interactome is not well known, but two recently identified homologous or heterologous interactors, Hsp104 and Aplysia CHIP are CC proteins (Si et al., 2003; F.F. and E.R.K., unpublished). Taking into account interactors of CPEBs listed in the IntAct database, 84% of the putative apCPEB interactors are CC proteins (Fig. 1B). Eukaryotic proteomes are estimated to contain only 6-8% CC proteins (Odgren et al., 1996). Thus, our findings indicate that CC proteins are overrepresented among the interactors of these Q/N-rich and polyQ proteins.
Our interactome findings prompted us to search for CCs in the Q/N-rich prions and polyQ proteins themselves, as potential mediators of interaction. Using Coils (Lupas et al., 1991) and Paircoil2 (McDonnell et al., 2006), two algorithms that detect CC heptad repeats in primary sequences, we analyzed six yeast Q/N-rich prions (Ure2, Sup35, Rnq1, Swi1, Cyc8, Mot3), the Aplysia Q/N-rich prion CPEB, and nine human proteins undergoing polyQ expansion (Htt, ataxin-1, -2, -3, -7, androgen receptor (AR), atrophin-1, calcium channel subunit α-1A (CACNA-1A), and TATA-box binding protein (TBP; Orr and Zoghbi, 2007; Patel et al., 2009; Alberti et al., 2009). We found that all of these proteins contain heptad repeats and are predicted with high probability (0.8-1) to contain CC domains that flank/overlap with Q/N-rich regions and polyQ stretches (Fig. 1B-C, S1).
Heptad repeats are within prion domains (PrDs) of the yeast and Aplysia proteins. In Ure2 and apCPEB, the CC region overlaps entirely with the regions defined as prion domains (Fig. 1B-C, S1C; Ure2 residues 1-95, Bousset et al., 2002; apCPEB residues 1-160, Si et al., 2003a). Swi1, Rnq-1, and Cyc8 display numerous regions of high CC propensity within their Q/N-rich domains (Fig. S1, S2). Sup35 shows CC propensity within the M section of its NM prion domain (residues 124-253; Krammer et al., 2008; Fig. S1, S2), which is crucial for prion propagation and Sup35 interactions with Hsp104 (Liu et al., 2002). Moreover, 80% of the 19 candidate yeast prions identified by Alberti et al. (2009) are predicted to have CC propensity (0.8-1, Paircoil2; Fig. 1B and S2).
All nine polyQ proteins show CC propensity in regions that contain polyQ stretches (Fig. 1B-C, S1). For some of them, the CC propensity of the wild-type (WT) polyQ regions is not high, but increases strongly upon polyQ expansion (Fig. S1B). For comparison, we analyzed prions and amyloids devoid of Q/N-rich or polyQ stretches, such as the mammalian prion protein (PrP), β-amyloid(1-42), tau and α-synuclein (Fig. 1C, S1). Except for α-synuclein, already known to form CCs (Bussell and Eliezer, 2003), these proteins do not show high CC propensity, suggesting that CC domains might represent a defining feature of the subgroup of Q/N-rich prions and polyQ amyloids. Fig. 1C and S1A show predictions for proteins with known CC crystal structures, including short CCs (c-Fos), long CCs (tropomyosin), and Q/N-enriched (>15%) CCs from the SIV protein gp41 and the Mycoplasma protein MPN010 (Yang et al., 1999; Shin et al., 2006).
We compared the features of predicted Q/N-rich and poly-Q CCs with known CCs (Fig. 2). Coiled-coil heptads are repeats of seven residues (a-b-c-d-e-f-g), in which hydrophobic residues (mostly L, I, V, M or F) often occupy positions a/d, creating a hydrophobic layer between coiling helices (Fig. 2A-C). Polar/charged residues are instead in solvent-exposed positions, and charged residues in e/g can form CC-stabilizing salt bridges. Repeats of hydrophobic residues with heptad spacing are a hallmark of CCs, and are often discontinuous (Parry et al., 2008, Fig. 2D, S2G). - Moreover, as seen in helical nets (e.g. gp41; Fig. 2D), hydrophobic residues in positions a/d may be interspersed with certain non-hydrophobic ones such as Qs, also considered ambivalent hydrophobes (Sodek et al., 1972). Q and N in position a/d may specify the oligomeric state of CCs (Gonzalez et al., 1996).
The putative Q/N-rich and polyQ CCs (Fig. 2D, S2G) contain a/d clusters of hydrophobes alternating mostly with Q/Ns, with few charged residues in e/g. Thus, predicted Q/N-rich and polyQ CCs have features comparable with known CCs, and suggest that they may be unstable and conformationally flexible (Li et al., 2003; Kwok and Hodges, 2004). A recurrent feature of these domains is the quite neat segregation of proline/glycine (P/G) residues outside of the CC-prone regions (Fig. S1C, S2D-E).
We studied further the CC domains of Ure2, apCPEB, and human Htt, the determinant of Huntington’s disease upon polyQ expansion. We found that the N-terminal portions of Ure2, apCPEB and polyQ-expanded Htt exon-1 (Htt-72Q), are identified as CCs by Paircoil2 and contain 10, 11, and 12 heptad repeats, respectively (Fig. S3A-B). Some heptads display a conventional amino acid composition, while others are quite atypical in this respect, being mostly formed by Q/N residues, either associated or not with hydrophobic residues in a/d. Orthologs of these proteins are also predicted to form CCs, suggesting that these structures are evolutionarily conserved because of their functional relevance (Fig. S3C-D).
To study the role of CCs in the aggregation and activity of Ure2, apCPEB and Htt exon-1, we used structure-guided mutagenesis to modulate the stability of putative CC domains. We designed several mutants (Fig. 3A) and estimated CC disruption/enhancement for each using Paircoil2 (Fig. 3B). In coiled-coil–defective (cc-) mutants, we substituted residues in a/d with single or tandem prolines (Fig. 3A), a commonly used strategy for CC disruption, impairing CC formation partially or completely (Chang et al., 1999; Fig. 3B). Given the neat separation of conventional and Q heptads in Htt-72Q, we used this protein as a suitable model to modulate CC propensity also in the opposite direction. We generated CC-promoting mutants (cc+) predicted to i) induce aggregation of the soluble Htt-25Q, ii) enhance Htt-72Q aggregation, and iii) rescue aggregation of cc- mutants. Thus, we replaced 4-6 Qs with hydrophobic residues in the a-d frame of heptads #1-2, thereby generating mutants Htt-25Q/cc+, Htt-72Q/cc+, and Htt-72Q/cc+R (Fig. 3A).
Since prolines in the cc- mutants may also affect β-sheet formation, we generated control mutants predicted to maintain propensity for CC (cc0), but in which the possible β-sheet formation would be impaired by replacing multiple Qs with β-sheet-breaking, charged or polar residues (Colloc’h and Cohen, 1991). We used glutamate (E), lysine (K), and asparagine (N), up to four each, in various combinations (Fig. 3A). To test the efficacy of these mutations in preventing β-sheet formation, we generated Aβ(1-42) mutants containing E, K, and N substitutions (β- mutants) with similar spacing as in the cc0 mutants (Fig. 3A). These mutations disrupted the β-strand propensity of Aβ(1-42) (Fig. S3E, S5B-C). As a further control for the proline mutagenesis, we generated the Htt-72Q/cc-/W mutant (Fig. 3A) in which four tryptophans (Ws) replaced the L-L-F-L pattern in a/d. Tryptophan in a/d is mildly CC destabilizing (Fig. S5E-F; Kwok and Hodges, 2004), but it has greater β-sheet propensity than leucine (L) (Pawar et al., 2007). Thus, this mutant would be expected to have preserved β-sheet-mediated aggregation but reduced CC-mediated aggregation.
To test the prediction that Q/N-rich and polyQ stretches can be part of α-helical CC structures, and to assess effects of the CC mutations, we used circular dichroism (CD) to analyze the secondary structure of three sets of peptides (Fig. 4A). CD is used extensively to study the folding and stability of CCs. Distinctive signatures permit differentiating between single and coiled helices, based on the ratio between 222 and 208 nm ellipticities (>1 for CCs), its inversion induced by trifluoroethanol (TFE), and the thermal stability of the folding (see Suppl. Experimental Procedures).
The first set of peptides is based on a four-heptad CC model (Hicks et al, 2002) into which we inserted two central heptads of Qs (peptide ccQQ), or Qs with either CC-stabilizing (I-L-I-L) or -destabilizing (P-P-P-P) residues in a/d (ccQL and ccQP, respectively). In benign buffer, ccQQ and ccQL displayed α-helical CD spectra with minima at 208 and 222 nm, whereas ccQP showed a minimum at ~200 nm as for random coils (Fig. 4B). Both ccQQ and ccQL had θ222/θ208 ratios >1, indicating CC formation. In 50% TFE, which favors CC dissociation, the θ222/θ208 ratio of ccQQ and ccQL went below 1 (Fig. 4C, S4A), as typically observed for CCs. TFE also induced helical folding of ccQP, likely by stabilizing its proline-flanking parts (Fig.S4A). Expected CC stabilization in ccQL vs ccQQ was realized; ccQL remained folded up to 75°C, whereas ccQQ lost most of its structure already at 25 °C (Fig 4D-E). We conclude that polyQ stretches can be part of CC structures, and that Q-rich CCs gain stabilization by hydrophobic residues with heptad spacing.
We similarly analyzed Ure2 and Htt peptides. Ure2(70-98), comprising heptads #7-10 (Fig. S3A), is only partially helical at 4°C, and becomes essentially unstructured above 25°C (Fig. 4F, S4B). Difference spectra (37°C vs. 4°C) revealed more clearly that the structured component is α-helical and coiled, as indicated by a θ222/θ208 ratio ~1 (Fig. 4F). This conclusion is also supported by the inversion of the ratio in 50% TFE, which also strongly enhanced α-helical folding (Fig. S4B). Htt-17Q (1-40), designed by reference to the only form of Htt that has been crystallized (Kim et al., 2009), generated CD spectra similar to Ure2(70-98), being only partially helical at 4°C, consistent with the crystal structure (Kim et al., 2009; Fig. S4C-D). Interestingly, unlike the other peptides, the CD spectra of Htt-17Q changed with time, showing a reduction in ellipticity at 208 nm and a progressively increasing θ222/θ208 ratio. After some hours at 4°C, the CD spectra showed a helical profile with θ222/θ208 >1, indicative of the CC formation, and moderate thermal stability (Fig. 4F-G). Subsequently, θ208 decreased further and the 222 nm minimum was red-shifted towards 225 nm (Fig. S4C-D); similar transitions have been observed in CD spectra of peptides forming CC fibers (e.g. Potekhin et al., 2001; Frost et al., 2005).
Finally, to assess the structural consequences of the CC-targeting mutations, we studied peptides derived from cc+ and cc- mutants. Htt25Q/cc+(1-31) displayed frank CC features (θ222/θ208 ratio > 1) and notable thermal stability (Fig. 4H-I). Conversely, Htt72Q/cc-/#2(1-31), comprising the first four heptads of Htt-72Q with prolines in a/d, displayed a random coil conformation (Fig. 4H).
Quantitative fittings for CD spectra of peptides with θ222/θ208 ratio > 1 (ccQQ, ccQL and Htt-25Q/cc+(1-31), Supplemental Table S1) confirm high helical content (40-75%) and show negligible levels of β-sheet structure (2-8%). Moreover, the thermal denaturation series (Fig. 4D) have isodichroic points at ~203 nm, typically found upon CC destabilization and indicative of an equilibrium between α-helices and random coils (Gazi et al., 2008). These results are in excellent accordance with predictions for Ure2 and Htt (Fig. 3B).
CC assemblies range from dimers to polymers (Parry et al., 2008). To define the oligomeric state of the peptides characterized by CD, we performed glutaraldehyde cross-linking experiments, which revealed dimeric and higher-order species (Fig. 4L,M,N). At 37°C, ccQL and Htt-25Q/cc+(1-31) are mostly in higher-order multimers; ccQQ forms supradimeric species as well, but less so than ccQL, consistent with its lower stability (Fig. 4E).
Non-cross-linked Htt-17Q(1-40) ran as dimers, although higher order forms were also present (Fig. 4M, arrow). Cross-linked Htt-17Q(1-40) generated an opalescent solution with visible aggregates indicative of large peptide assemblies, which were not able even to enter the gel (Fig. 4M, white asterisk). These and CD results demonstrate the marked polymerization tendency of Htt-17Q(1-40).
Ure2(70-98), which is marginally folded at 4°C, did not display significant formation of supradimeric species at 37°C (not shown), in accord with CD. The proline-containing peptides ccQP and Htt-72Q/cc-/#2(1-31) were mostly monomeric, displaying only a very modest degree of dimerization (Fig. 4L,N).
These findings show that α-helical CC-prone polyQ and Q/N-rich peptides form higher-order oligomers in vitro, and indicate the possibility that CCs may trigger protein aggregation in vivo.
To determine the relevance of CCs for in vivo aggregation, we compared the subcellular distribution of WT and cc- mutant Ure2, apCPEB and Htt-72Q overexpressed as GFP fusions.
WT and cc- Ure2 were expressed in a knock-out yeast strain (ure2Δ) to prevent any influence of endogenous Ure2. Ure2-GFP formed aggregates in numerous cells (25.7 ± 1.1%, n=36 50×50 μm microscopic fields of cultures from 3 colonies; Fig. 5A), whereas the cc- mutant had mostly diffuse subcellular distribution, forming aggregates in significantly fewer cells (p<0.01, t-test; Fig. 6E).
We next overexpressed WT or cc- mutant apCPEB in Aplysia neurons (Fig. 5B). ApCPEB formed aggregates in the soma, rapidly decreasing in number along the axon, and no diffuse fluorescence was detectable between them, or in distal neuritic branches (Fig. 5B, 1-4). Conversely, mutant cc-/#2 presented diffusely along the neurites (Fig. 5B, 7-8), down to distal branches where no aggregate was detectable. Mutant cc-/#1, predicted to disrupt CC only partially (Fig. 3B), had an intermediate phenotype (data not shown). The cc-/#2 mutant still formed aggregates proximally to the cell body, similar to what observed on overexpressing truncated apCPEB devoid of its Q/N-rich domain (Si et al., 2010). These results are similar to what is found for a non-Q/N-rich RNA-binding protein (PABPN-1), whose aggregation relies both on an N-terminal CC and a C-terminal RNA-binding domain (Tavanez et al., 2005).
Finally, we compared the distribution of Htt-72Q and its cc- mutants in HEK293 cells 72h after transfection (Figs. 5C). Aggregation was significantly impaired for mutant cc-/#1, and almost completely abolished for cc-/#2. ANOVA (Fig. 6F) showed an overall effect of the CC mutations on aggregation (F(12,337) = 103.11, p<0.001), with significant differences between Htt-72Q and both its cc- mutants (p<0.01, Newman-Keuls test). These phenomena were not cell-type-specific (Fig. 5D, S5A). In addition, we tested the role of CCs in the heterotypic interactions of Q/N-rich and polyQ proteins. We found that CC fragments of interactors are recruited into aggregates (Fig. S6A-B) and that CC destabilization impairs the Htt-72Q/CHIP interaction (Fig. S6C). These findings demonstrate a close correlation between CC propensity of Ure2, apCPEB, and Htt-72Q and their aggregation properties.
We next analyzed the aggregation of CC-enhancing (cc+) mutants. Htt-25Q does not aggregate upon overexpression, but the Htt-25Q/cc+ mutant formed multiple aggregates per cell in a proportion of cells not different from the Htt-72Q group (Fig. 6A, F), showing that a non-aggregating form of Htt can be induced to aggregate by increasing its CC propensity, independently of polyQ expansion. The possibility that this aggregation was a consequence of β-sheet formation induced by the additional hydrophobic residues is ruled out by the CD results (Fig. 4H-I). Furthermore, we found that the addition of the same L-L-F-L pattern with heptad spacing amidst the polyQ stretch of Htt-72Q (Htt-72Q/cc+) increased aggregation (p<0.01), and that a similar pattern plus an additional F-L pair (Htt-72Q/cc+R) was able to substantially rescue the aggregation of cc-/#1 (p<0.001; Fig. 6A, F).
These experiments show that increasing the CC propensity of Q stretches can induce or enhance aggregation in vivo, and they support the notion that Q stretches may be part of CCs, especially when stabilized by flanking or overlapping heptad repeats of hydrophobic residues.
To rule out that the effect of proline substitutions in cc- mutants may be due to β-sheet disruption rather than CC destabilization, we studied the effect of mutations disfavoring β-sheet but not CC formation (cc0 mutants), and compared the effect of the same mutations on the aggregation of Aβ(1-42), a known β-sheet-forming amyloid (β-mutants).
All cc0 mutants formed aggregates after 72 hours of overexpression (Fig. 6B) in ~50-60% of the cells (i.e. 70-85% of the Htt-72Q aggregation rate), a significantly higher proportion than for cc- mutants (p<0.01, Newman-Keuls test). Interestingly, the microstructure of cc0 aggregates became more elaborate as the number of substitutions increased, showing spiny protrusions, star-like figures, and eventually tangled fibers (Fig. 6D).
On the other hand, while GFP-Aβ(1-42) aggregated in many of the overexpressing cells (17.71 ± 2.51%, n=20 fields; 391 cells), the aggregation of β- mutants #1 and #2 was completely abolished (p<0.01; Fig. 6C, S5D), thus demonstrating the efficacy of the cc0 mutations in disrupting β-sheet-based aggregation. Furthermore, Htt-72Q/cc-/W formed smaller aggregates than Htt-72Q (Fig. 5C, S5E-G), as expected for CC-driven aggregation since L→W substitution is mildly CC destabilizing but β-sheet enhancing.
These findings further support the notion that the aggregation of Q-rich proteins relates essentially to their CC propensity and that conversion to β-sheets may not be crucial for triggering aggregation.
Aggregated amyloids display detergent insolubility, which can be assayed by ultracentrifugation of cell lysates to separate detergent-insoluble aggregates from soluble forms. We collected lysates of yeast and HEK293 cells overexpressing Ure2 and Htt variants for 72 hours. Given the lack of mass transfection systems for Aplysia cells, we expressed apCPEB variants in HEK293 cells, which were lysed 12-18 h after transfection, given the faster aggregation kinetics of this protein in these cells.
WT Ure2, apCPEB, and Htt-72Q were detectable in both soluble and insoluble fractions, (Fig. 7A,C,E). However, all cc- mutants displayed a remarkable reduction of the detergent-insoluble fraction (Fig. 7B,D,F). The data for Htt mutants were normalized to Htt-72Q (65.6 ± 2.4% in the pellet fraction, n=13), and revealed an overall significant effect of the mutations on solubility (F(10,55)=33.95, p<0.001; one-way ANOVA Fig. 7F). The insoluble fraction of the cc-mutants was strongly reduced with respect to Htt-72Q (p<0.001, Newman-Keuls test). Conversely, the proportion of the cc+ mutant in the pellet was significantly increased as compared with Htt-72Q (p<0.03), and, the cc+R mutant showed a rescued insolubility with respect to cc-/#1 (p<0.001). Finally, cc0 mutants were found in the pellet fraction in a proportion >80% of that of Htt-72Q, significantly more than cc- mutants (p<0.001).
Thus, the detergent insolubility correlated with CC propensity, closely paralleling the aggregation phenotypes. These findings are also consistent with other observations of detergent-insoluble structures from CC proteins (e.g. Yang et al., 2002).
We next sought to determine whether CC disruption also interferes with Ure2 function. We used ureidosuccinate (USA) uptake to monitor the presence of [URE3] in Ure2Δ cells overexpressing WT or mutant Ure2, grown under plasmid selection on either uracil or USA substrate (Fig. 7G-H). Both WT and mutant transformants grew equally well on the non-selective uracil-containing substrate, whereas cc- transformants grew significantly less on USA. The cc- transformants formed about half of the colonies formed by WT Ure2 transformants, as determined at dilutions from 10-3 to 10-5 (51.2 ± 2.5%, n=18 vs 100 ± 4.6%, n=30, p<0.001, t-test; Fig. 7H). These results indicate that the N-terminal CC structure of Ure2 is important not only for aggregation but also for prion induction.
The cytotoxicity of polyQ-expanded proteins can be assayed in vitro; thus, we tested the impact of CC mutations on the activity of Htt-72Q using a colorimetric assay to detect the cellular reduction of MTT to formazan. We observed significant cytotoxicity in HEK293 cells after 72h of Htt-72Q overexpression; formazan production was significantly less after Htt-72Q transfection than in mock-transfected control cultures (i.e. 80.4 ± 1.8% in n=45 culture wells versus 100 ± 1.5% for n=46 controls, p<0.001 t-test). Then, we compared the Htt-72Q cytotoxicity with that of its CC mutants, normalizing the toxicity of each mutant to that of Htt-72Q (Fig. 7I). ANOVA revealed a significant overall effect of the mutations on Htt-72Q toxicity (F(11,402)=13.10, p<0.001). The cc- mutants were the only two Htt-72Q mutants devoid of toxicity, whereas all the other mutants retained substantial toxicity. The toxicity of cc-/#1 and /#2 was not different from the control group (p>0.42). The cc+R mutant was significantly toxic with respect to control (p<0.04), as well as the cc+ mutant (p<0.001). The cc0 mutants all displayed significant toxicity, ranging between 42 and 63% of the Htt-72Q toxicity (p<0.02).
These results show a strong correlation between the CC propensity of Htt-72Q variants and their toxicity, similar to that we observed for their aggregation and insolubility. The essential role of the non-polyQ heptads in mediating Htt-72Q dysfunctional activity is evident in the fact that their disruption alone (cc-/#1) was sufficient to abolish toxicity.
We have identified CC domains as a recurrent feature of Q/N-rich and polyQ proteins. The α-helical CC supersecondary structure mediates a variety of functions, including protein polymerization and conformational change (Parry et al., 2008). Based on our experiments, we further suggest that the known properties of CCs are well suited for explaining some of the unique features of Q/N-rich prions and polyQ amyloids. In particular, regulated CC-based structural transition may account for the physiological conformational transitions of functional Q/N-rich prions better than stochastic misfolding mechanisms.
We found that CC domains are overrepresented in Q/N-rich and polyQ proteins and in their interactomes, and that CC fragments of polyQ interactors are recruited into Htt-72Q aggregates. Moreover, CC disruption impairs CHIP-Htt-72Q interactions. These findings strongly suggest that CC interactions underlie the recognition between Q/N-rich and polyQ domains and their interactors. The importance of CCs in chaperones has been recognized previously (Martin et al., 2004), and “intrinsically disordered” domains can form complex CC interactomes with refined functional properties (Gazi et al., 2008). Interestingly, the chaperone TRiC suppresses Htt aggregation by binding to the initial 17 Htt residues (N17; Tam et al., 2009), which we show to be CC-stabilizing. In turn, CCs deletion in proteasomal subunits impairs their recruitment into polyQ aggregates (Rousseau et al., 2009). Our results show that these CC domains are indeed sufficient to induce the recruitment of fused GFP into Htt aggregates. CC interactions may thus play a fundamental role in the propagation of Q/N-rich prions and polyQ toxicity, processes that depend on interactions with chaperones and other proteins.
Q/N-rich and polyQ domains are generally considered as low-complexity, disordered structures prone to form β-sheets, similar to conventional amyloids (Wickner et al., 2007; Orr and Zoghbi, 2007). Our findings show however that these domains not only display α-helical features but also have propensity to form CCs, highly organized supersecondary structures with functional properties. Alternative structures have been proposed for polyQ segments (e.g. Thakur et al., 2009; Kim et al., 2009), and our findings are consistent with studies indicating the α-helicity of Q/N-rich or polyQ domains (Lathrop et al., 1998; Bousset et al., 2002; Leitgeb et al, 2007; Davies et al., 2008; Kim et al., 2009). We detected heptad repeats flanking or overlapping Q/N-rich and polyQ domains, and displaying features of unstable CCs (Li et al., 2003). In fact, these CCs contain variable associations of canonical and atypical heptads, which might define the relative stability of each CC.
Our CD experiments support the conclusion that polyQ heptads are compatible with α-helical CC structures. CD also demonstrated how Q/N-rich or polyQ-containing fragments of Ure2 and Htt form helical structures in vitro that display signatures of CCs and, in the case of Htt-72Q(1-40), of CC fibers. The in vitro stability of the Htt CC is enhanced by hydrophobic stabilizers, and can be disrupted by a few prolines. These observations validate the mutational strategies that we adopted for the in vivo experiments.
The notion that α-helical Q/N-rich and polyQ domains can form CC structures is also supported by crystal structures of non-amyloid proteins that contain CCs relatively enriched with Q/N residues (e.g. Yang et al., 1999; Shin et al., 2006; Guo et al., 2007). Interestingly, recent studies have identified the polyQ-flanking N17 sequences of Htt as amphipathic α-helical mediators of nucleation for Htt oligomers (Kelley et al., 2009; Tam et al., 2009), in excellent agreement with our findings (although these studies consider such dynamics as a prelude to the polyQ misfolding into β-sheets). Furthermore, Tam et al. (2009) recognized the importance of the hydrophobic face of N17, which we identify as an essential CC-stabilizing element, for Htt aggregation and interactions with TRiC. At variance, Thakur et al. (2009) propose a different mechanism whereby the unfolding of a scarcely structured N17 induced by polyQ expansion leads to β-sheet nucleation; however, N17 was invariably α-helical in Htt-17Q crystals (Kim et al., 2009). Finally, Williamson et al. (2010) proposed that oligomerization of polyQ-expanded Htt occurs via contacts between polyQ stretches in a collapsed globular conformation, forming micellar structures in which β-sheets assemble. In this model, N17 represents a suppressor of the intrinsic polyQ associativity, a conclusion that runs counter to Tam et al. (2009) and to our data. In agreement with our and other studies, however, the same study highlights the helical propensity of N17 and its role in mediating interactions.
Our findings strongly implicate CCs as mediators of Q/N-rich and polyQ protein aggregation. In fact, cross-linking experiments show that Q-rich CC peptides in vitro can form dimers and higher-order multimers, indicating that they may drive aggregation in vivo. To test this possibility in vivo, we generated mutants of Q/N-rich and polyQ proteins in which the CC propensity was modulated predictably and in opposite ways. CC disruption hampered the aggregation of Ure2, apCPEB and Htt-72Q, whereas the mutants with enhanced CC stability (cc+) strikingly demonstrated how non-aggregating polyQ stretches can be induced to aggregate by the addition of few CC stabilizing residues. These findings are consistent with the established role of CCs in protein polymerization and fibrillogenesis (Parry et al., 2008). CCs can mediate aggregation (Fig. 7J and S7B) through side-by-side bundling, or intermolecular swapping of helices (e.g. Deng et al., 2007; Ogihara et al. 2001).
β-Sheet-breaking residues did not abolish Htt-72Q aggregation (cc0 mutants) but they did disrupt Aβ(1-42) aggregation, indicating that a transition to β-sheet is not required to trigger polyQ aggregation. This finding suggests that α-helical CCs might be self-sufficient mediators of the aggregation of these proteins; however, we cannot rule out the possibility that CCs are an intermediate step (Abedini and Raleigh, 2009) whereby β-sheet formation might occur as a consequence of an aggregation process initially driven by CCs (Narayanan et al., 2006; Xu, 2007; Abedini and Raleigh, 2009). CCs and β-sheets may also co-exist in some instances in which aggregation might be mediated by distinct domains with different structural features (Fig. S7A).
Our findings also implicate CCs in the pathogenesis of polyQ-expansion diseases. All nine disease-related polyQ proteins contain putative CCs, and CC disruption impairs Htt-72Q toxicity. In particular, we found that heptads #1-2 (N17) are important not only for aggregation, as recently observed (e.g. Tam et al., 2009), but also for Htt toxicity, and may be a trigger region to drive CC folding of the poly-Q stretch. This interpretation rationalizes in structural terms the known context-dependency of the pathogenicity of polyQ stretches (Orr and Zoghbi., 2007) by explaining how a flanking region may promote polyQ aggregation. The CC nature of polyQ domains may also explain the puzzling pathogenic threshold in polyQ length. Various polyQ diseases appear when Q stretches exceeds a critical length of about 35 residues and severity progressively increases with polyQ expansion (Orr and Zoghbi, 2007). The pathogenic threshold corresponds to the fifth Q heptad and more heptads generate longer CC-prone helices. These extensions may acquire a stronger polymerization tendency, arising from easier side-to-side interactions among longer coils and/or from the swapping between protomers via the added polyQ heptads. Dysfunctional protein-protein interactions of polyQ-expanded proteins are major determinants of the polyQ disease neuropathology (Orr and Zoghbi, 2007). Most transcription factors sequestered by polyQ proteins contain CCs, as do molecular motors and synaptic SNAREs. Degenerate CC interactions, resulting from non-specific coiling of polyQ-expanded helices with CC-prone helices of captured proteins, could represent a fundamental determinant of the impairment of transcription and other cellular processes, leading ultimately to cell death.
The Ure2 mutant with CC-destabilizing residues has an impaired capability to induce the [URE3] prion, which shows that the CC structure is involved not only in the aggregation but also in the biological activity of Ure2, and raises the possibility that the CC might be the prionic conformation. The fact that CC regions are present within the prion domains of known Q/N-rich prions strongly indicates that they can play an important role in the structural dynamics of these proteins. The mechanisms of the conformational switch of Q/N-rich prions and other aspects of their biology are still quite obscure (Wickner et al., 2007).
Prions are thought to undergo a conformational switch to a self-perpetuating conformation. If this conformation is a CC, then the prion switch can be conceived either as a folding from random-coil to α-helical CC or as a rearrangement of helices or pre-assembled CCs (Fig. 7J). Both possibilities are compatible with known properties of CCs, and they can be physiologically regulated by a variety of mechanisms, including post-translational modifications (Fig. S7B). The self-perpetuating prionic conformation can be conceived of as a self-templating CC, capable of promoting the folding or coiling of other protomers, seeding their polymerization/aggregation. The structural constraints of such processes may be responsible for the species barrier between prion orthologs. Prion curing could occur when a protein, e.g. a chaperone, sequesters prion monomers through CC binding, making them unavailable for homopolymerization. Similarly, CC-based structural transitions could explain how prion proteins with functional properties like apCPEB can switch between functional states through self-sustaining conformational changes, triggered in a regulated manner by specific signaling events (Si et al., 2010) rather than by the stochastic misfolding of protomers as in conventional models of prionogenesis.
A CC-based model does not exclude the interplay of CCs with other secondary structures, including β-sheets, in the conformational transitions of Q/N-rich prions, and it can be reconciled with conventional models of the prion switch. Notably, the NM prion domain of Sup35 (Krammer et al., 2009) contains both a CC region within the M section and a Q/N/G/Y-rich region with no CC propensity within the N section, which has similarity with the β-sheet-prone mammalian prion PrP (Liu and Lindquist, 1999). Thus, some prion proteins can combine features of both yeast Q/N-rich and mammalian non-Q/N-rich prions. α-Helical structures can drive Sup35 multimerization, and β-sheet formation may occur only after initial aggregation (Narayanan et al., 2006). In principle, α-helical and β-sheet structures in different protein domains could also act in parallel to mediate aggregation. The aggregation of apCPEB in vivo seems to rely both on the Q/N-rich prion domain and to some extent also on the C-terminal RNA-binding domain (Si et al., 2010). Consistent with this view, parallel studies of purified apCPEB carried out in vitro support the possibility of a complex interplay of α-helical, random coil, and β-sheet structures in the formation of catalytically active prion fibers (Raavendra, Siemer, Hendrickson, Kandel and McDermott, unpublished data). Such complexity and heterogeneity in the conformational transitions of Q/N-rich prions is also suggested by sequence analyses that define at least three sub-classes of prion domains (Q/N-rich, Q/N/P/G rich, P/G-rich) (Fig. S2). Thus, the combinatorial interplay of structurally heterogeneous domains may ultimately define the specific aggregation and activation pathways of each individual prion protein (Figs. 7 and S7).
In conclusion, our findings indicate that coiled-coil domains can have a key role in the aggregation and function of a subset of prions and amyloids enriched in Q/Ns residues, and indicate that structural mechanisms underlying the switch of Q/N-rich prions and polyQ amyloidogenesis can be quite heterogeneous. Within the conventional β-sheet model, CCs might be regarded as intermediate structures in the transition of helical domains to β-sheets (Fig. 7J, red arrows; Abedini and Raleigh, 2009), or as facilitators of β-sheet formation in neighboring regions (Fig. S7A). On the other hand, our experiments indicate the possibility of a novel, alternative model for the Q/N-rich and polyQ subset of prions and amyloids, in which CCs are self-sufficient mediators of the prion switch and amyloidogenesis (Fig. 7J). This mechanism may have important implications for the distinction between pathogenic and functional prions. Whereas the toxicity of pathogenic prions and amyloids relies on a misfolding into β-sheets, non-pathogenic and functional Q/N-rich prions may use CC-based conformational changes for switching physiological activity on and off in a regulated and persistent manner. Some non-pathogenic Q-based CCs may acquire pathogenic features when genetic mutations cause abnormal extensions. From this perspective, unstable or multistable coiled-coils with tendency to homopolymerize - physiologically present in some functional proteins and pathologically extended in some others – may underlie the epigenetic transmission of prion traits in yeast, the long-term storage of information at synapses, and the pathogenesis of devastating molecular diseases.
Interactomes were derived from IntAct and BioGrid databases (www.ebi.ac.uk/intact/; www.thebiogrid.org/). The algorithms Coils (Lupas et al., 1991) and Paircoil2 (McDonnell et al. 2006) were used for CC predictions. Coils was used for interactome screenings, and both programs were used for subsequent analyses. For graphical uniformity with the Coils plots, the graphs of Paircoil2 predictions in the figures display the per-residue CC propensity as 1 minus the P-score assigned to each amino acid as an estimate of CC probability.
Site-directed mutagenesis was performed using the QuikChange Multi Site-directed Mutagenesis kit (Stratagene).
Peptides were synthesized by GenWay Biotech (San Diego, CA), dissolved in benign buffer (100mM NaCl, 20 mM phosphate buffer, pH 7.5) or in the same buffer with 50% TFE (Sigma), and CD spectra were acquired with AVIV 202SF or AVIV 400 spectropolarimeters.
Peptides in benign buffer were incubated at 37 °C for 15 min with 0.05% glutaraldehyde. The reaction was stopped by adding 0.1M ethanolamine, samples were run on tricine-SDS-PAGE gels and silver-stained.
A S.cerevisiae MATα Ure2 knockout strain (ure2Δ) was obtained from Open Biosystems (Clone 11983). Aplysia neurons were cultured following previous protocols (Si et al., 2003). Yeast and HEK293 cells were maintained following standard procedures.
Yeast transformation and HEK293 transfection were performed using the S.c. EasyComp kit (Invitrogen) and Fugene-6 (Roche), respectively. Plasmid DNA was microinjected into Aplysia neurons.
Cells were imaged with an Olympus FV500 confocal scanning system.
HEK293 cells were lysed 12-72 hours after transfection. Lysates were centrifuged and proteins in the pellet and supernatant were detected by Western blotting with an anti-GFP antibody (Clontech). Band intensities were quantified using ImageJ (NIH).
HEK293 cells were transfected with Htt constructs, and after 72 hours, cell viability was assessed by a MTT-formazan colorimetric assay. Samples were analyzed spectrophotometrically with a microplate reader (Labsystems Multiskan MS).
We thank Emi Ling for technical help with HEK cells, Francisco Monje, Burkhard Rost and Ann McDermott for helpful discussions and suggestions, and Craig Bailey, Christoph Kellendonk and Radegonda Mazzitelli for critical reading of the manuscript.
Detailed experimental procedures, seven figures and one table can be found with this article online.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.